Abstract
2296
Objectives Software packages of automated method for volumes of interest (VOIs) are widely used. However, we reported (JNM 52: 2045) that tested automated methods achieved <90% spatial agreement against our manual method (MM) and resulted in suboptimal agreement of binding potential (BP) and amphetamine-induced dopamine release (DArel) measured with [11C]raclopride. Thus, we propose and evaluate a supervised-automated method (SM).
Methods VOIs for putamen (Pu), caudate nucleus (CN), and cerebellum (Cb) were defined on MRIs of 20 healthy subjects who had baseline and post-amphetamine [11C]raclopride scans by MM and SM, and using Freesurfer (FS; Fischl et al., 2004), FSL/FIRST (FF; Patenaude et al, 2011), and SPM/unified segmentation Method (SU; Ashburner et al., 2003). In MM, VOIs were defined initially using low/high thresholds and manual elimination of continuations to surrounding structures (Tx-mode), and refined meticulously in three orthogonal views. In SM, VOI slices defined by the Tx-mode on every three trans-axial images were interpolated in z-direction and smoothed by a 3mm FWHM Gaussian kernel. Striatum VOIs were divided in to anterior (a) and posterior (p) subdivisions (e.g., aPu), and to ventral striatum (vS; Own vS for FS and FF). BP was obtained by MRTM2 (Ichise et al., 2002) and DArel was calculated as one less amphetamine-baseline BP ratio.
Results The manual portion (Tx-mode) of SM ranged from 1.5 min for CN to 4 min for Cb. Spatial agreement (common/total) in MRI space, deviations of baseline BP and DArel from MM are listed in Table.
Conclusions Proposed SM outperformed tested automated VOI methods in spatial agreement and on baseline BP. Thus, SM appeared to be a promising method, especially being far less time consuming than MM. The findings warranted further evaluation of SM in larger samples including various neuropsychological conditions, and wider structures and receptor/transporter systems.
Research Support NIDA, NIAAA, NIM